摘要
基于分块的压缩感知算法适用于图像信号的处理,通过平滑迭代阈值投影法可以快速重构图像,但存在低采样率下重构图像质量较差的缺点。基于全变差分的分块压缩感知算法,在一定程度上能提升重构效果,但降低了运算速度。针对以上算法的不足,提出基于多尺度的自适应采样图像分块压缩感知算法。根据小波分解后不同层对重构结果影响所占权重不同的特性,自适应分配给每一层不同的采样率,并在重构时将平滑迭代阈值投影法应用到每一层的每一个子带的分块上。实验结果表明,与传统的迭代阈值投影法相比在重构质量上提高了1~3 d B,在重构速度上与迭代阈值投影法相当并优于全变差分法。
The block based compressive sensing algorithm can be applied to the image signal processing, and the image can be reconstructed quickly by using the method of smoothed projected landweber. Due to the algorithm display a poor reconstruction quality under the low sampling rate, someone propose the algorithm based on the total variation, which is called TV algorithm. Despite a certain improvement was made on the reconstruction effect, the algorithm decreased the operation speed on the other hand. In view of the deficiency of the two algorithms, we pro- pose an adaptive sampling block based compressive sensing algorithm based on multiple scales. According to the difference in wavelet decompo- sition layers on the reconstruction results, we adaptively alloeat each layer of different sampling rate, and apply the Smoothed Projected Land- weber algorithm to each layer of each sub zone block. Experimental results reveal that the proposed algorithm improves the reconstruction quali- ty by one to three dB, and the method is superior to the total variation method in the reconstruction speed .
出处
《微型机与应用》
2016年第24期42-45,49,共5页
Microcomputer & Its Applications
基金
国家自然科学基金(61170102)
关键词
压缩感知
多尺度
小波变换
自适应采样
图像分块
compressed sensing
multiscale
wavelet transform
adaptive sampling
image blocking
